A doctor knows that barometric pressure reliably drops before her patient's migraines. She tells the patient: 'Your migraine occurred because the barometer fell.' Is this a scientific explanation?
AYes — a reliable statistical correlation between two events is sufficient for scientific explanation
BYes — because the barometer allows prediction of the migraine, it provides a full explanation of it
CNo — the doctor needs to cite a law of nature explicitly; correlation without a stated law is not explanatory
DNo — the barometer reading is a correlate of the atmospheric change, not the cause; a genuine explanation would cite the actual causal process linking atmospheric pressure to neural events
The barometer falls because atmospheric pressure drops — the same change that (by some causal pathway) triggers the migraine. The barometer reading and the migraine share a common cause; neither causes the other. Citing the barometer 'explains' the migraine no more than citing the migraine could explain the barometer falling. This is the core insight: explanation requires the right kind of connection — causal or structural — to what produced the event, not merely a reliable pattern. Prediction can succeed on correlation alone; explanation cannot.
Question 2 Multiple Choice
By the deductive-nomological (DN) model, a valid explanation must deduce the explanandum from laws of nature plus initial conditions. What is the 'asymmetry problem' that challenges this model?
AThe DN model applies only to deterministic laws, making it unable to handle probabilistic explanations in quantum mechanics or biology
BThe DN model requires too many initial conditions, making it practically impossible to apply to complex real-world events
CThe DN model licenses explanations in both causal directions — shadow length could 'explain' flagpole height just as validly as flagpole height explains shadow length
DThe DN model conflates explanation with description, since both involve citing facts about the world
The flagpole/shadow case illustrates the asymmetry problem precisely. The flagpole's height + sun angle entail the shadow length (a genuine explanation). But the shadow length + sun angle also mathematically entail the flagpole height — a valid DN 'explanation' that feels explanatorily backwards. We know the flagpole causes the shadow, not vice versa, but the DN model has no way to encode this asymmetry because it is purely logical. The explanatory direction is determined by causation, not by deductive validity — which is why causal theories of explanation were developed in response.
Question 3 True / False
According to the DN model, predicting an event before it occurs and explaining it after it occurs require different logical structures.
TTrue
FFalse
Answer: False
The DN model explicitly holds that prediction and explanation have the same logical structure. Both involve deriving a statement about an event from laws of nature plus initial conditions. The only difference is temporal: in prediction, you derive the event's occurrence before it happens; in explanation, you derive it after. Hempel called this 'the structural identity thesis.' Critics argued this was counterintuitive — it implies that any successful prediction is, in principle, also an explanation — but it follows directly from the DN model's purely logical account of explanation.
Question 4 True / False
The asymmetry problem for the DN model shows that logical deducibility alone is insufficient to distinguish genuine explanations from explanatorily irrelevant or backwards derivations.
TTrue
FFalse
Answer: True
This is precisely the lesson of the flagpole example and related cases (like explaining a person's height from their shadow, or explaining why a patient took aspirin by deriving it from a law that aspirin relieves headaches plus the fact that the patient had a headache — but their taking aspirin explained the headache disappearing, not vice versa). Valid logical entailment from laws and conditions is necessary but not sufficient for explanation. What's missing is directional constraint — causal, mechanistic, or otherwise — that the purely formal DN model cannot supply.
Question 5 Short Answer
What is the difference between predicting that an event will occur and explaining why it occurred? Use a concrete example to show why successful prediction does not guarantee genuine explanation.
Think about your answer, then reveal below.
Model answer: Prediction requires only a reliable correlation or pattern: knowing the barometer falls before storms lets you predict rain without knowing why. Explanation requires identifying what actually produced the event — typically its cause, mechanism, or the law-governed structure responsible for it. A barometer predicts rain but does not explain it; atmospheric physics (low pressure systems causing moisture to condense) explains rain. Similarly, knowing a patient's age and risk factors may let a doctor predict a heart attack, but the explanation requires understanding the causal pathways of cardiovascular disease.
This distinction is central to philosophy of science. Science aims at both prediction and explanation, but they are not the same achievement. A purely predictive science — even one with perfect accuracy — would leave us without understanding of why the world works as it does. The puzzle the DN model was trying to solve is that explanation seems to involve more than correlation, and more than prediction: it tracks something about the underlying structure of reality — causes, mechanisms, or deep laws — that prediction does not require.